Automation audit
Where your hours go, what is worth automating, and what to leave alone.
Fixed quote, in writing
Hand off the work that repeats. Keep the decisions.
We build automations for businesses in Quebec and across Canada that run on their own, stop for your approval where a decision is yours, and are built to alert a person when a step fails. AI does part of the work; a person reviews it and answers for what we deliver.
- Runs on its own1. Quote request arrivesAn email or web form comes in.
- Runs on its own2. Details pulled outName, job, address and deadline go into your quote template.
- Runs on its own3. Draft quote preparedYour prices, your wording.
- Waits for a person4. You approve or editThe quote does not go out until you approve it.
- Runs on its own5. Sent and filedThe client gets it, your records are updated.
Alert path, from steps 2 and 5
Alerts someoneIf a step fails, a person is alerted.Built so a failed step is flagged, not buried.
- Runs on its own
- Waits for a person
- Alerts someone
How we design every system to behave
Runs on its own
Routine steps run without anyone watching.
Waits for a person
Steps that commit you (money, a promise, a message to a client) are set to wait for your approval. A step runs on its own only if you agree to it in writing.
Alerts someone
Each system is built to alert a person when a step fails, instead of failing quietly.
What we build
Workflow automation
The recurring work nobody has time for, rebuilt to run on its own.
Fixed quote, in writing
AI workspace setup and training
Claude set up on your own computer, with training until you run it alone.
Setup quote, hourly training
Assistants and agents
For work that has to read, decide and act, with your approval on the actions we agree need it.
Fixed quote, in writing
Run and improve
Monitoring, adjustments and re-tests when the model underneath changes.
Flat monthly fee
Agent-ready website
A machine-readable layer that gives AI assistants accurate facts about you to work from.
Fixed quote, in writing
What we build with
You keep the workflow files, the prompts and your own accounts.
- Claude/LLM
- Gemini/LLM
- n8n/automation
- MCP/tools
- Supabase/data
- Cloudflare/infra
- Netlify/infra
- DigitalOcean/infra
- Resend/email
- GitHub Actions/ci
- ElevenLabs/voice
- Zoho/biz
Product names are trademarks of their owners. Naming a tool means we use it, not that its vendor partners with or endorses us.
Three typical jobs
Every job is quoted in writing before work starts: a fixed price for the scope of a build or setup, an hourly rate for training, a flat monthly fee for ongoing care. Third-party accounts are yours and are listed in the quote.
Quotes drafted for your approval
- Your email or web form as the trigger
- Draft quotes from your price list and wording
- Your approval before anything is sent
Fixed price, quoted in writing
Invoices and follow-ups without the chasing
- Invoices issued from finished jobs
- Reminders at the intervals you choose
- You decide anything past the first reminder
Fixed price, quoted in writing
Claude set up and taught on your own computer
- Installed under your own account
- Two or three tools built around your repeat work
- Hands-on training until you run it alone
Setup quoted in writing, training by the hour
- Drafted
- With AI assistance
- Reviewed and approved
- A person
- Sheet
- 1/1
Who answers for it
Every inquiry is read by a person, and every deliverable is reviewed by a person before it goes out. AI research agents draft the field notes and help with parts of client work; they are named on this site. Ask, and we will tell you where they are used on your work.
What gets checked
- Scope and price, in writing, before any work starts
- That every approval step works before a system goes live
- The test results again whenever the model underneath changes
- Cost per step, monthly, for the builds we monitor on a retainer.
Where your data goes
Your AI provider account is yours. Any account we host or operate for you is named in the quote. Before a build starts, you get this sheet filled in for your system, to use in your own Law 25 records. It supports your privacy impact assessment; it does not replace it.
| Provider | What it processes | Where it is processed |
|---|---|---|
| AI model provider | The text a step sends it | Named in writing |
| Workflow platform | The records the workflow moves | Named in writing |
| Email sending | Messages the workflow sends | Named in writing |
Field notes
All field notesOur own measurement
Your AI instructions have a shelf life. Audit them before they fail.
Illustrative example
AI agents will use your website or route around it. WebMCP bets on the first.
Illustrative example
Your AI is stuck. Restart the stronger model, don't hand it the transcript.
Questions
Do I have to get on a call?
No. Describe the task in writing. The first read, the scope and the quote all come back in writing. A call happens only if you ask for one.
Where does my data go?
You keep your own account with the AI provider, so the account, the billing and the provider's terms are yours. We use the smallest data that does the job and tell you in writing which providers process it and in which locations, so you can document it for your own privacy obligations. Some providers process data outside Quebec; we tell you which. Running a model on your own hardware is possible: it is a compliance choice, not a saving.
Do I own what you build?
Yes, as set out in your agreement: the workflow definitions, prompts and documentation we write for you are yours, and we hand them over as files with handover notes. The third-party tools they run on (n8n, your AI provider, hosting) stay under their vendors' licences and accounts. Documented for handover: the notes are written so another team can pick the work up.
What happens when the models change?
They change often, and instructions written for an old model can make a new one worse. On a retainer we re-test your setup when the model it runs on changes and adjust it. Without one, you keep the checks and the documentation.
What does it cost?
Every engagement is quoted in writing before work starts: a fixed price for the scope in the quote on setup and builds, an hourly rate for training, a flat monthly fee for ongoing care. Scope changes are quoted before they are done. Third-party subscriptions (your AI provider and any tool accounts the build needs) are yours, billed to you directly, and listed in the quote.
Do I need an AI agent?
Often, no. A form, a rule or a simple assistant handles most recurring work, costs less and fails less. We start with the simplest thing that works and move up only when the task needs judgment across several steps.
How do you keep it from making things up?
We build it to work from your documents rather than from the model's memory, and to mark missing facts instead of filling them in. Outputs are checked against a fixed test set when the build changes, and a person approves the steps we agree need it. This reduces invented content; it does not eliminate it, which is why the approval step stays.
